I tested performance of pnfs by running iozone in eight threads for sequential write and sequential read and compare it the nfsv4. Test is performed in 6x2 and 3x1 volumes.

1.) Create a 6x2 , 3x1 gluster volumes
2.) Export using nfs-ganesha
3.) Mount the volume using pnfs and nfsv4
4.) run iozone for file size varying from 64MB to 8GB on both mounts
iozone -Rwc  -i {0,1} -r 64k -s <file size>  -t 8 

All the servers and clients are connected using 1Gb ethernet card(max possible speed 128MB/s)

Results :
Write performance is better than nfsv4, Gap is more on distributed volumes
Read performance is fluctuating(may be more or equal or even less). Also there are iozone read errror in the iozone logs.

Created attachment 1052543 [details]
Performance_result

Correcting last sentence of my previous comment:

And also there are read errors in the iozone logs

moving this BZ to verified.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://rhn.redhat.com/errata/RHSA-2015-1495.html